Model Pricing and Billing

Core billing unit: The platform bills per 100 million Tokens. A 1x multiplier = 40 yuan per 100 million Tokens, equivalent to 0.4 yuan per 1 million Tokens. The “multiplier” and “price per 1 million Tokens” columns in the table below use the same conversion basis, so you can read the prices directly from the table.

One Key works with every model. Multipliers only represent differences in model costs. You do not need a separate Key for each model.

Actual charges are based on system records. If a model receives a temporary subsidy, promotional discount, or multiplier adjustment, the price table on this page will be updated accordingly.

What is a cache hit? When consecutive requests contain a large amount of identical context, the system can reuse previously processed content. The cached portion does not need to be recomputed at the full input rate. Common examples include long conversations, consecutive follow-up questions, code completion, agent workflows, and multi-turn tool calls.

Whether a request results in a cache hit depends on the request content, model capabilities, upstream caching policy, and current system availability. Final charges are based on billing records.

How to read the table: Start with the model name, then check the multiplier, and finally review the “price per 1 million Tokens.” A lower multiplier means a lower cost per Token. Multipliers and prices use the same conversion basis.

API endpoint: Use https://api.llm-token.cn/v1 whenever possible. If a client is incompatible with an address that includes /v1, try https://api.llm-token.cn instead. Different API gateway platforms, clients, and plugins may have different Base URL requirements. After switching models, we recommend testing with a small request first.

The price table change monitoring Skill has been updated to 1.0.4. The new version fixes issues reading large tables in public Feishu documents and prioritizes the actual model price table, preventing explanatory tables from being misidentified as price tables.

How to use it: The first run creates a baseline. Run it on a schedule after that. An output of NO_REPLY means the price table has not changed. When changes are detected, it automatically generates a notification covering added models, removed models, multiplier changes, and price changes.

Scope: This Skill is designed for publicly accessible price tables in Feishu documents. For private documents or pages that require authentication, verify access permissions and data security requirements before connecting them.